do-task
Pass
Audited by Gen Agent Trust Hub on Aug 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process user-provided specification files located in
docs/specs/task-UCS/. These files act as instructions for subagents to generate source code and tests. This creates a surface for indirect prompt injection if the specifications contain malicious instructions aimed at influencing the subagent's code generation or execution behavior.\n - Ingestion points: The skill reads all markdown files within the
docs/specs/task-UCS/directory during the orchestration process.\n - Boundary markers: There are no explicit delimiters or warnings provided to the subagents to ignore potentially malicious instructions embedded within the specification files.\n
- Capability inventory: The skill grants subagents the ability to write source code files to the project directory and execute system commands for building and testing (e.g.,
go build,npm run build,cargo build).\n - Sanitization: The skill does not implement any validation or sanitization of the contents of the specification files before they are processed by the subagents.\n- [DYNAMIC_EXECUTION]: The skill facilitates the automated generation of executable code and the subsequent invocation of build tools. This dynamic behavior is the intended primary purpose of the skill to support backend development workflows and does not represent a security violation in this context.\n- [COMMAND_EXECUTION]: The skill instructs subagents to execute standard build and test commands (e.g.,
go build,npm run build,cargo build) within the project environment to verify the generated code.
Audit Metadata